This application is in response to RFA 90-CA-05, which involves the Radiologic Diagnostic Oncology Group (RDOG) multi-institutional collaborative imaging trials. A team of radiologists, surgeons, pathologists and a statistician has been assembled to assess the relative accuracies of computed tomography (CT) and magnetic resonance (MR) imaging in the preoperative staging of primary malignant neoplasms of the musculoskeletal system. One hundred consecutive new patients with nonvisceral soft-tissue sarcomas of the pelvis and extremities, and 40 consecutive new patients with malignant primary bone neoplasms of the pelvis and extremities will be enrolled in this study each year for four years at the Memorial Sloan-Kettering Cancer Center. State-of-the-art CT and MR scans will be obtained in random order on each patient within two weeks before surgery. The preoperative CT and MR scans for each patient will be interpreted independently by separate radiologists without access to the clinical, surgical, or pathologic data; detailed data forms will be completed by the radiologists at that time. A similar data form will be completed by the surgeon at the completion of the operation, and by the pathologist at the time of gross and microscopic evaluation. The pathologic findings, amplified by surgical findings and by subsequent clinical and radiologic followup, will constitute the "gold standard" against which the imaging findings will be compared. The data, aggregated as well as stratified by tumor location and type, will be analyzed using receiver operating characteristic (ROC) curves.